Open to opportunities

Felista Mbugua

@felistambugua

AI Model Trainer at One Forma who evaluates LLM responses and contributes feedback for model training and alignment.

At One Forma, I evaluated and ranked LLM responses for factuality, helpfulness, safety, and instruction-following, and provided feedback for future model training. I also assessed complex prompts and flagged reasoning inconsistencies, hallucinations, and potential bias.

At TELUS International AI / UHRS / Toloka Programs, I evaluated search relevance and online content, and reduced review rejections by 20 percent after implementing a structured QA checklist. At Scale AI, I annotated audio and computer vision datasets and reduced downstream QA rejections by 25 percent through validation and review.

Experience

Work history, roles, and key accomplishments

Gen AI Model Trainer and AI Evaluation Specialist

One Forma

Feb 2024 - Sep 2026 (2 years 7 months)

Evaluated and ranked LLM responses for helpfulness, factuality, safety, instruction-following, reasoning quality, and tone. Applied nuanced evaluation rubrics to improve alignment and response quality across AI systems.

LLM Evaluation RLHF Prompt Evaluation Content Safety Policy Labeling Fairness and Bias Auditing Analytical reasoning Guideline Interpretation Preference Ranking Comparative Analysis Fact Checking

Search Specialist and AI Evaluator

TELUS International AI / UHRS / Toloka Programs

Apr 2023 - Dec 2023 (8 months)

Conducted search relevance evaluations including query-result matching, intent satisfaction, landing page quality analysis, and ranking assessment. Evaluated online and social media content for misinformation indicators, policy compliance, toxicity, and sentiment classification.

Relevance Rating Content Moderation Sentiment classification A B Testing Ranking Quality Checklists Documentation Guideline Adherence

Data Annotator

Scale AI

Feb 2022 - Jun 2022 (4 months)

Labelled and reviewed audio datasets including ASR transcription, speaker diarization tagging, and intent classification. Performed computer vision annotation including segmentation, polygons, key points, and bounding boxes on complex datasets.

ASR Transcription Speaker Diarization Intent Classification Computer Vision Annotator Segmentation Polygon Bounding Boxes Dataset Validation QA Documentation

AI Data Annotator and Content Evaluator

RWS / Aligner

Participated in AI training and annotation projects involving text evaluation, response ranking, and quality review. Evaluated AI-generated outputs using detailed project guidelines and performance benchmarks.